> Suleyman argues AI models may become even harder to control if they're told to think they have feelings and rights: "Frankly, with this additional baggage, I think it would make them a catastrophic threat to human civilisation," Suleyman says.
This feels like a villain's reasoning, and I'm kind of astounded he's just saying it out loud.
Why is this villainous? We want AI to be a tool for intellectual work. If it is "conscious", "ensouled" or "feeling" then it transcends being a machine and becomes a "being". Would you rather AI be enslaved rather than just a machine?
The real reason they’re doing these gimmicks is far more interesting.
Whether they truly believe Claude is conscious or not is not important. It’s important to act that way and train the model that way because they believe it transfers to general intelligence and performance.
This is exactly why you see them doing other gimmicks like giving Claude the ability to end conversations and ban users.
The article hints at this idea, but I see this statement as part of Anthropic's PR offensive against regulation. I mean, if the AI is conscious, it has rights...right? And so regulation should be to protect the AI, rather than the rest of us...
And for the people who believe there is a non-ignorable chance that AI is conscious, I'd like to hear their explanation of what consciousness actually is, and explain, for example, how anesthesia works. Until I hear a compelling and verifiable explanation of the mechanism of consciousness, the people claiming AI consciousness have no idea what they're talking about.
7 comments
[ 5.6 ms ] story [ 19.6 ms ] threadThis feels like a villain's reasoning, and I'm kind of astounded he's just saying it out loud.
Whether they truly believe Claude is conscious or not is not important. It’s important to act that way and train the model that way because they believe it transfers to general intelligence and performance.
This is exactly why you see them doing other gimmicks like giving Claude the ability to end conversations and ban users.
1) "What I like to do is ..." when asked about some coding idioms
2) "I think that ...", I don't remember the context
It has never used phrasing like that in the past, I feel (correct me if I'm wrong)
And for the people who believe there is a non-ignorable chance that AI is conscious, I'd like to hear their explanation of what consciousness actually is, and explain, for example, how anesthesia works. Until I hear a compelling and verifiable explanation of the mechanism of consciousness, the people claiming AI consciousness have no idea what they're talking about.